On Wed, Mar 22, 2017 at 06:19:12PM +0100, Jernej Škrabec wrote:
Hi,
Dne sreda, 22. marec 2017 ob 08:45:48 CET je Maxime Ripard napisal(a):
On Tue, Mar 21, 2017 at 11:26:46PM +0100, Jernej Škrabec wrote:
Hi,
Dne torek, 21. marec 2017 ob 20:34:33 CET je Maxime Ripard napisal(a):
Hi,
On Mon, Mar 20, 2017 at 11:01:25PM +0100, Jernej Skrabec wrote:
diff --git a/include/configs/sun50i.h b/include/configs/sun50i.h index 1b7bfb6c22..146f7f4e1b 100644 --- a/include/configs/sun50i.h +++ b/include/configs/sun50i.h @@ -21,6 +21,8 @@
#define GICD_BASE 0x1c81000 #define GICC_BASE 0x1c82000
+#define CONFIG_SUNXI_DE2
/*
- Include common sunxi configuration where most the settings are
*/
diff --git a/include/configs/sun8i.h b/include/configs/sun8i.h index a4c3fb69e4..c42b901107 100644 --- a/include/configs/sun8i.h +++ b/include/configs/sun8i.h @@ -25,6 +25,10 @@
#define CONFIG_SUNXI_USB_PHYS 2
#endif
+#ifdef CONFIG_MACH_SUNXI_H3_H5 +#define CONFIG_SUNXI_DE2 +#endif
/*
- Include common sunxi configuration where most the settings are
*/
diff --git a/scripts/config_whitelist.txt b/scripts/config_whitelist.txt index 8e5dc36fa7..ba0eb12665 100644 --- a/scripts/config_whitelist.txt +++ b/scripts/config_whitelist.txt @@ -3102,6 +3102,7 @@ CONFIG_STV0991_HZ_CLOCK
CONFIG_ST_SMI CONFIG_SUN4 CONFIG_SUNXI_AHCI
+CONFIG_SUNXI_DE2
CONFIG_SUNXI_EMAC CONFIG_SUNXI_GMAC CONFIG_SUNXI_GPIO
This should be a Kconfig option.
So hidden option in board/sunxi/Kconfig will probably be the best then?
Yes, I guess, but I'm not entirely sure why you need two different options there?
I used define for CONFIG_SUNXI_DE2 here because SoC has or has not DE2 IP block and that is not a choice. Option in patch 5 is configurable and gives opportunity to build video driver or skip it, whithout influencing clock structure.
I could make only one option, but then it would have to be configurable, which doesn't really make sense from patch 4 perspective, because, as I already stated before, this is property of the SoC.
Which solution do you prefer? One option, define and option (as it is now) or two options?
Frankly, none of them is ideal. Best solution would be to convert clocks to use driver model framework.
I guess you could make a hidden Kconfig option selected by the relevant MACH_ options.
As a general basis, we move away from the old-style config options, so adding any new !Kconfig options isn't really an option.
Maxime